<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="wang-etal-2026-probing">
    <titleInfo>
        <title>Probing the Think–Say Gap on Sensitive Concepts in Chinese and English Open LLMs</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Yu-Chieh</namePart>
        <namePart type="family">Wang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Lang-Ching</namePart>
        <namePart type="family">Yeh</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Guan-Yu</namePart>
        <namePart type="family">Tseng</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Shu-Kai</namePart>
        <namePart type="family">Hsieh</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2026-10</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>The Proceedings of the Workshop on Open Reasoning Across Cultures Languages (ORACLE)</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Leonardo</namePart>
            <namePart type="family">Ranaldi</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Giulia</namePart>
            <namePart type="family">Pucci</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Federico</namePart>
            <namePart type="family">Ranaldi</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Marco</namePart>
            <namePart type="family">Valentino</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Alexandra</namePart>
            <namePart type="family">Birch</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Fabio</namePart>
            <namePart type="given">Massimo</namePart>
            <namePart type="family">Zanzotto</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">Budapest, Hungary</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
        <identifier type="isbn">979-8-89176-483-5</identifier>
    </relatedItem>
    <abstract>Open large language models (LLMs) trained under different alignment regimes treat politically sensitive concepts (freedom, democracy, human rights) differently, but outputs alone cannot separate alignment-induced suppression from pretraining culture: a model may withhold what it internally represents, or may simply reproduce the distribution it was trained on. We compare a Chinese-developed model (Qwen2.5-7B-Instruct) with an Englishdeveloped baseline (Gemma-3-12B-IT) on 360 bilingual, minimally contrastive items with a country-swap control (China vs. a matched democracy, Germany), using Natural Language Autoencoders to verbalize hidden states so that models with incommensurable representation spaces can be compared through text. Both models linearly encode a sensitive-vsneutral direction at the sentence-final state, beyond emotional salience and in both languages. Their behavior, however, diverges sharply: the Chinese model is China-specifically suppressive, justifying restrictions on political grounds far more for China than for a matched democracy, while the English model shows the reverse. The internal side remains inconclusive. Read before the model answers, the state does not support a reliable judgment of whether critical content is present, so we do not claim a think–say gap and identify extraction during generation as the decisive next test.</abstract>
    <identifier type="citekey">wang-etal-2026-probing</identifier>
    <location>
        <url>https://aclanthology.org/2026.oracle-main.20/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>304</start>
            <end>315</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Probing the Think–Say Gap on Sensitive Concepts in Chinese and English Open LLMs
%A Wang, Yu-Chieh
%A Yeh, Lang-Ching
%A Tseng, Guan-Yu
%A Hsieh, Shu-Kai
%Y Ranaldi, Leonardo
%Y Pucci, Giulia
%Y Ranaldi, Federico
%Y Valentino, Marco
%Y Birch, Alexandra
%Y Zanzotto, Fabio Massimo
%S The Proceedings of the Workshop on Open Reasoning Across Cultures Languages (ORACLE)
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-483-5
%F wang-etal-2026-probing
%X Open large language models (LLMs) trained under different alignment regimes treat politically sensitive concepts (freedom, democracy, human rights) differently, but outputs alone cannot separate alignment-induced suppression from pretraining culture: a model may withhold what it internally represents, or may simply reproduce the distribution it was trained on. We compare a Chinese-developed model (Qwen2.5-7B-Instruct) with an Englishdeveloped baseline (Gemma-3-12B-IT) on 360 bilingual, minimally contrastive items with a country-swap control (China vs. a matched democracy, Germany), using Natural Language Autoencoders to verbalize hidden states so that models with incommensurable representation spaces can be compared through text. Both models linearly encode a sensitive-vsneutral direction at the sentence-final state, beyond emotional salience and in both languages. Their behavior, however, diverges sharply: the Chinese model is China-specifically suppressive, justifying restrictions on political grounds far more for China than for a matched democracy, while the English model shows the reverse. The internal side remains inconclusive. Read before the model answers, the state does not support a reliable judgment of whether critical content is present, so we do not claim a think–say gap and identify extraction during generation as the decisive next test.
%U https://aclanthology.org/2026.oracle-main.20/
%P 304-315

Markdown (Informal)

[Probing the Think–Say Gap on Sensitive Concepts in Chinese and English Open LLMs](https://aclanthology.org/2026.oracle-main.20/) (Wang et al., ORACLE 2026)

• Probing the Think–Say Gap on Sensitive Concepts in Chinese and English Open LLMs (Wang et al., ORACLE 2026)

ACL

• Yu-Chieh Wang, Lang-Ching Yeh, Guan-Yu Tseng, and Shu-Kai Hsieh. 2026. Probing the Think–Say Gap on Sensitive Concepts in Chinese and English Open LLMs. In The Proceedings of the Workshop on Open Reasoning Across Cultures Languages (ORACLE), pages 304–315, Budapest, Hungary. Association for Computational Linguistics.